Abstract:Data augmentation is a powerful technique to enhance the performance of a deep learning task but has received less attention in 3D deep learning. It is well known that when 3D shapes are sparsely represented with low point density, the performance of the downstream tasks drops significantly. This work explores test-time augmentation (TTA) for 3D point clouds. We are inspired by the recent revolution of learning implicit representation and point cloud upsampling, which can produce high-quality 3D surface reconstruction and proximity-to-surface, respectively. Our idea is to leverage the implicit field reconstruction or point cloud upsampling techniques as a systematic way to augment point cloud data. Mainly, we test both strategies by sampling points from the reconstructed results and using the sampled point cloud as test-time augmented data. We show that both strategies are effective in improving accuracy. We observed that point cloud upsampling for test-time augmentation can lead to more significant performance improvement on downstream tasks such as object classification and segmentation on the ModelNet40, ShapeNet, ScanObjectNN, and SemanticKITTI datasets, especially for sparse point clouds.

What problem does this paper attempt to address?

The paper aims to address the issue of Test-Time Augmentation (TTA) for 3D point cloud data. Specifically, the paper focuses on the problem where the performance of downstream tasks (such as object classification and semantic segmentation) significantly degrades when 3D shapes are represented by low-density point clouds. To solve this problem, the authors propose a method based on implicit representation learning and point cloud upsampling to enhance the data during testing. The main contributions of the paper are as follows: 1. Analyzing and comparing existing reconstruction methods, including surface-based sampling and point cloud upsampling, for test-time data augmentation. 2. Proposing a test-time augmentation method suitable for 3D point cloud deep learning, which applies to two different strategies: implicit representation and point cloud upsampling. 3. Discovering a self-supervised point cloud upsampling method as a robust test-time augmentation approach, utilizing near-surface cues to sample enhanced point clouds. 4. Demonstrating the effectiveness of their augmentation method on two downstream tasks, including object classification and semantic segmentation, through extensive experiments and analysis, showing significant performance improvements on both synthetic and real-world datasets. The core idea of the paper is to treat the input point cloud as a noisy estimate of the underlying surface and propose generating additional point clouds through implicit fields or point cloud upsampling techniques to enhance the data during testing. These enhanced point clouds can be used for downstream tasks such as classification and segmentation to improve overall performance.

Test-Time Augmentation for 3D Point Cloud Classification and Segmentation

Test-Time Adaptation in Point Clouds: Leveraging Sampling Variation with Weight Averaging

Test-Time Adaptation of 3D Point Clouds via Denoising Diffusion Models

Test-time augmentation for deep learning-based cell segmentation on microscopy images

TauPad: Test Data Augmentation of Point Clouds by Adversarial Mutation

S$^3$-TTA: Scale-Style Selection for Test-Time Augmentation in Biomedical Image Segmentation

Real3D-Aug: Point Cloud Augmentation by Placing Real Objects with Occlusion Handling for 3D Detection and Segmentation

PointAugment: an Auto-Augmentation Framework for Point Cloud Classification.

TauLiM

Improved Text Classification via Test-Time Augmentation

What Can be Seen is What You Get: Structure Aware Point Cloud Augmentation

Test-Time Augmentation Meets Variational Bayes

TauPad

Performance Enhancement Using Data Augmentation of Point Cloud Based 3D Object Detection for Autonomous Driving

Improving 3D Object Detection through Progressive Population Based Augmentation

PointAugmenting: Cross-Modal Augmentation for 3D Object Detection

TTT-KD: Test-Time Training for 3D Semantic Segmentation through Knowledge Distillation from Foundation Models

TauLiM: Test Data Augmentation of LiDAR Point Cloud by Metamorphic Relation

Quantifying Data Augmentation for LiDAR based 3D Object Detection

Boosting Anomaly Detection Using Unsupervised Diverse Test-Time Augmentation

Efficient Improvement of Classification Accuracy via Selective Test-Time Augmentation